In the fifth installment of our GC×GC Detective Stories series, Philip Marriott – together with Renée Webster and Konstantinos Kouremenos – explores three cases in which comprehensive two-dimensional gas chromatography revealed chemical information hidden by conventional separations.
The investigation begins with the search for a biomarker associated with a rare newborn metabolic disorder, where a diagnostically important compound was concealed beneath a much larger chromatographic peak. It then moves to aviation fuel, where trace oxidation products had to be distinguished from an overwhelmingly complex hydrocarbon matrix. The final case concerns “shape-shifting” molecules that interconvert during chromatographic separation, producing a mysterious plateau whose composition could not be resolved using one-dimensional GC.
Marriott and colleagues take us through the three investigations and explain what the additional separation dimension allowed the researchers to discover.
Story 1: The Biomarker Hidden in Plain Sight
Newborn babies are routinely screened for various genetic disorders. Most people will be familiar with the heel-prick sampling performed soon after birth. There is now an incredible library of blood spot samples stored in hospitals.
Analytical procedures are fundamental to assessing these disorders and may draw on a broad range of techniques, depending on the markers associated with the particular condition. These markers can range from large molecules, such as proteins, to small molecules, such as short-chain acids. The appropriate technique will depend on the molecule, but identifying the marker to target is no easy task.
Considerable research may be needed to identify the disease, determine whether specific markers can connect a disease or symptom with the endpoint of a metabolic pathway, and establish how best to isolate and measure those markers. The analyst must consider the sample matrix, extraction procedures, and the sensitivity required for the measurement.
Here, we outline a discovery strategy using GC×GC to search for new small-molecule markers of aciduria – an inborn error of metabolism (IEM).
Inborn errors of metabolism are mostly monogenic disorders caused by defects in enzymes or transporters that disrupt a step in an intermediary pathway. When flux through the normal route stalls, the substrate and its proximal precursors accumulate and are shunted into secondary reactions, including β-oxidation, ω-oxidation, transamination, glycine or carnitine conjugation, and glucuronidation.
Many of the resulting metabolites are carboxylic acids that exceed renal reabsorption thresholds and spill into urine. Diagnosis rests on detecting the direct substrate of the enzyme or a downstream by-product. Examples of species monitored in IEMs include methylmalonic acid; 3-hydroxyisovaleric acid; glutaric and 3-hydroxyglutaric acids; ethylmalonic, adipic, and suberic acids; orotic acid; succinylacetone; and N-acetylaspartate.
Choosing the right analytical approach
Methodologies have generally evolved in the following order:
Colorimetric or spot tests, including DNPH, cyanide-nitroprusside, and Benedict’s tests
Paper and thin-layer chromatography
GC-MS of derivatized urinary organic acids
FIA-MS/MS of acylcarnitines and amino acids in dried blood spots, which forms the basis of expanded newborn screening
LC-MS/MS for the same analytes and for more polar or labile species
GC-MS remains the reference technique for urinary organic acid analysis because organic acids are small, chemically diverse, and amenable to derivatization to increase their thermal stability and volatility. The GC-MS workflow includes:
Forming oxime derivatives of keto acids using hydroxylamine
Acidifying the sample, saturating it with NaCl, and extracting into ethyl acetate or diethyl ether
Silylating hydroxy and carboxy groups with BSTFA or MSTFA to yield TMS ethers and esters
LC-MS/MS can also be used for this task. It avoids derivatization, and many analytes are not suited to GC. However, it also has some limitations relative to GC. Universal spectral libraries comparable to those available for electron ionization do not exist for LC-MS, so identification largely depends on accurate mass, MS/MS, and reference standards. Matrix effects, including ion suppression, can be substantial in urine. Very small, highly polar acids, such as lactate, glycolate, and oxalate, may also elute near the void on reversed-phase columns. Most metabolic laboratories therefore employ GC-MS and LC-MS/MS in parallel to cover complementary chemical space.
Revealing the hidden biomarker
Using GC×GC-TOFMS, we identified crotonylglycine as a novel urinary biomarker of 3-hydroxy-3-methylglutaryl-CoA synthase 2 (HMGCS2) deficiency. The analyte had eluded detection by conventional one-dimensional GC-MS because it coeluted with a substantially more abundant adipic acid peak. Separation in the second dimension resolved the two components, exposing the underlying diagnostic signal.
Elevated crotonylglycine was subsequently confirmed in further patients with proven HMGCS2 deficiency. However, the urinary organic acid profile can be transient and is most clearly manifested during episodes of metabolic decompensation, making the deficiency difficult to identify at other times.
A later, expanded patient study further evaluated crotonylglycine as a more persistent and diagnostically informative urinary marker of HMGCS2 deficiency. Its susceptibility to false-negative results when samples are collected during periods of metabolic stability underscores the importance of sampling during, or shortly after, an acute presentation.
This case illustrates a wider paradigm in which advances in chromatographic resolution have expanded the biomarker repertoire available for diagnosing IEMs. The HMGCS2/crotonylglycine example is one of several instances in which a clinically relevant marker is obscured in one-dimensional GC-MS. Other examples include:
Methylmalonic and succinic acids, which are notoriously difficult to resolve fully using single-column systems. Misintegration can affect quantification when concentrations are elevated, whereas further separation on a polar second-dimension column resolves the pair.
Ethylmalonic and methylsuccinic acids in short-chain acyl-CoA dehydrogenase deficiency, ethylmalonic encephalopathy, and multiple acyl-CoA dehydrogenase deficiency.
Story 2: Following the Biofuel Oxidation Trail
High-performance fuels, such as those used in military fighter jets, are specialized products that must conform to tight tolerances and performance criteria. When the fuel powering a precision aircraft degrades, the consequences can be catastrophic.
This was borne out in a little-known incident in 1976, when US Navy pilot Commander P. W. Ogle was forced to eject from his A-7E Corsair after engine failure during a flight from the aircraft carrier USS Ranger. The subsequent investigation attributed the failure to oxidized jet fuel. Although the incident occurred 50 years ago, oxidative degradation remains an important challenge for modern high-performance aviation fuels, for which chemical stability is essential to reliable operation.
During routine transportation, storage, and use, modern military and specialty jet fuels are exposed to air and temperature fluctuations that can induce oxidation reactions. The resulting products can form gums and deposits, attack fuel-system components, and ultimately compromise engine performance.
Aviation fuel is a highly complex material that may contain hundreds of thousands of different chemical species. Against this background, the oxidation products capable of producing such dramatic downstream effects are present only at trace concentrations and are difficult to resolve from the dominant hydrocarbon matrix. As new bio-based and other alternative fuels enter the market, they too will be subject to similar oxidative stress.
Finding trace products in a complex matrix
Assessing trace oxidation products requires either extracting them from the fuel or developing a method for analyzing them in situ, followed by an instrumental method capable of identifying the products. Although this might sound like a straightforward matter of applying an effective extraction strategy – for example, solvent extraction (SE), solid-phase extraction (SPE), or solid-phase microextraction (SPME) – the biofuel matrix makes effective removal of the matrix while retaining the oxidation products difficult. In some cases, a range of extraction and instrumental methods has failed to isolate and accurately identify the oxidation products in alternative fuel samples.
The target products are likely to include acids, alcohols, ketones, aldehydes, and compounds such as furanones. It is also important to measure these compounds at trace levels so that the early stages of degradation can be detected and mitigated. Measuring antioxidants added to a sample – even one as complex as a mineral fuel or biofuel – is relatively straightforward using targeted analysis and an MS/MS strategy. This can indicate when an added antioxidant is approaching depletion, but it does not address the presence of oxidation by-products, which remain a primary target of the analysis.
A separation strategy for oxidation products
Our research group offered an industry workshop on advanced multidimensional gas chromatography (MDGC) and comprehensive two-dimensional gas chromatography (GC×GC). Participants were invited to bring samples – or, ideally, send them to us before the workshop – so that we could develop an analytical strategy tailored to the sample and the aims of the analysis.
For a biofuel sample supplied by one participant, we wanted to measure the full range of oxidation products. GC×GC appeared ideally suited to resolving polar compounds within a non-polar matrix. The reverse problem – resolving non-polar compounds in a polar matrix – would also be well suited to GC×GC. Our experience with petrochemical analysis, and with the chemical-class speciation it provides for alkanes, saturated cyclic compounds such as cyclohexanes and cyclopentanes, olefins, and aromatic compounds, suggested several possible strategies.
Of the two broadly classified GC×GC column-set options – non-polar/polar (NP/P) or polar/non-polar (P/NP) – the most obvious choice was P/NP. With this configuration, alkanes are the most strongly retained analytes on the second-dimension column. The biofuel matrix consisted primarily of non-polar compounds, particularly alkanes. We therefore established conditions under which the alkanes eluted within an acceptable second-dimension retention time (²tR), as determined by the modulation period (PM). The polar, oxygenated compounds then eluted earlier. This is advantageous because the high concentrations of alkanes do not interfere with the earlier-eluting polar compounds. By contrast, with an NP/P column set, very polar compounds may exceed the PM setting and undergo wrap-around, while earlier-eluting, highly abundant alkanes may tail and cause interference.
During the workshop, a single GC×GC analysis revealed the full complexity of the oxygenated products. At a given first-dimension retention time (¹tR), the retention order was acids < alcohols < aldehydes ≈ ketones. The sheer number of compounds was astounding.
No extraction procedure – which might have selectively extracted or lost different oxygenated species – was required. GC×GC was exceptionally well suited to this analytical problem. Its speed, selectivity for oxygenated species, sensitivity, and ease of interpretation through the 2D separation space elution patterns made it a powerful approach. The method could be used to investigate changes over time or under different experimental and environmental conditions, including variations in storage, oxygen exposure, temperature, and humidity, without requiring extensive and costly extraction procedures.
We also developed a novel sampling, trapping, oven-cooling, and temperature-programming approach to achieve still greater resolution when needed. This involved heartcutting a very narrow, 12 s portion of the biofuel elution from the first column, cryogenically trapping the fraction at the head of the second-dimension column, cooling the oven, and then separating the fraction on the second-dimension column using a second temperature-programmed analysis.
The resulting strategy brought together the principal features of a successful analytical study, with GC×GC firmly at its center.
Story 3: The Curious Case of Shape-Shifting GC Molecules
In 1984 at the National University of Singapore (NUS), Professor Lai Yee Hing and author (PM) supervised an honors student, Ms Tan. As part of her project, she synthesized 9,10-bis(2-methylphenyl)phenanthrene.
The GC spirit in me suggested that we should study the GC analysis of one of her intermediates, which we expected to be semivolatile. That first product and analysis, at about 300 °C, produced a remarkable and fascinating phenomenon.
With packed-column GC, we observed a shoulder on the main peak – an unusual result. Seeking better separation, we moved to capillary GC. The “pure” compound now produced two well-resolved peaks. More remarkably, the peaks appeared to be joined by a raised plateau, or bridge.
Discussing the phenomenon with Professor Lai, we presumed that the two discrete peaks corresponded to E and Z isomers associated with the positions of the 2-methylphenyl substituents at the 9- and 10-positions of the phenanthrene. The peaks had different abundances, suggesting differences in the energetics of forming the E and Z isomers. The plateau appeared to arise from an on-column, sterically hindered rotational isomerization that converted E to Z and Z to E. At the high oven temperature, the energy supplied during GC analysis must have been sufficient to promote this isomerization.
Changing the oven temperature and flow rate altered the plateau in the way expected for a kinetic process. We then synthesized an anthracene analogue with 2-methylphenyl substituents at the 1- and 2-positions. In this case, the E and Z isomers had very similar abundances, giving the overall peak-and-plateau profile an almost symmetrical shape.
Similar observations were made for 2-methylnaphthalene with Cr(CO)₃ π-bonded to one ring. The chromium group can migrate between the rings, giving two resolvable terminal peaks connected by a plateau. We also observed the phenomenon for chromium tris(trifluoroacetylacetonate), which has Δ and Λ forms and again produced a plateau.
Much later, we became aware of other on-column structural rearrangements, including Cope rearrangements and the different plateau shapes observed in myrrh oil. Another example is the conversion of pyrethrum to isopyrethrum in chrysanthemum extracts, involving on-column conjugated double-bond shifts during GC analysis. Much of our work, however, has focused on E/Z isomerism in oxime compounds, R-C(H)NOH. Their lower elution temperatures are accompanied by correspondingly lower energy requirements for isomerization.
The problem of the plateau
We had observed the plateau, but still lacked a satisfactory explanation of how the separation process produced it.
The plateau between the two terminal peaks also presents a particular identification problem. In most cases, the two compounds are isomers – usually E/Z isomers – making it impossible to use strategies such as GC-MS to quantify the individual species within the plateau. The ¹D separation is intriguing, but the composition of the plateau cannot be quantified.
By this time, however, we had developed cryogenic modulation for GC×GC. It therefore seemed logical to apply GC×GC to this shape-shifting process.
Revealing the hidden profiles
GC×GC makes it possible to measure the actual amounts of the two isomers at any point over the plateau. It can also address more complex cases, including the enantioselective separation of chiral oximes.
Provided the individual compounds – in this case, oxime isomers – can be separated on the ²D column, the complete profile can be resolved within the ²D separation space. There should be no difficulty in achieving this because GC×GC is designed to do exactly that.
Importantly, the same stationary phase can be used for both the ¹D and ²D columns. The reasoning is straightforward: if compounds A and B can be resolved using a particular phase on the ¹D column, they should remain resolvable using the same phase when they coexist throughout the plateau region. This is subject to a sufficient difference in retention factor on the short ²D column, while the very narrow input band entering that column should further promote separation.
Using the same stationary phase for the ¹D and ²D columns still constitutes GC×GC. Provided that the ²D column resolves species that remain unresolved on the ¹D column, the method meets the requirements of an effective GC×GC separation.
The hidden profiles of the individual species can then be clearly revealed. We have reported the analysis of a racemic mixture of oximes, with enantiomer separation on the ¹D column, as well as the unprecedented separation of the enantiomers on both the ¹D and ²D columns. This is shown in Graphic 3; the corresponding ¹D GC result in panel A is difficult to interpret.
The curious case of on-column interconversion therefore provides a clear example of how multidimensional separation can reveal information that is inaccessible using ¹D GC.
Philip J. Marriott is Emeritus Professor of Chemistry at Monash University, Australia; Renée L. Webster is a research chemist at Commonwealth Scientific and Industrial Research Organisation (CSIRO), Australia; and Konstantinos A. Kouremenos is a senior researcher at NutriPATH Integrative and Functional Pathology Services, Australia
This article is part of our GC×GC Detective Stories series, exploring how comprehensive two-dimensional gas chromatography is being used to unravel analytical mysteries. Explore more from the series:
The Case of the Disappearing Diesel
Newsletters
Receive the latest analytical science news, personalities, education, and career development – weekly to your inbox.
